Description
When running a Kind (Kubernetes in Docker) cluster on OrbStack, the Kind nodes are unable to reach any external HTTPS endpoint including Docker Hub (registry-1.docker.io). This prevents the cluster from pulling container images at runtime.
DNS resolution works correctly through OrbStack's internal resolver (0.250.250.254), but TCP/HTTPS connections timeout or fail with TLS errors.
Environment
- OS: macOS (Apple Silicon / arm64)
- OrbStack Docker Engine: 29.4.0
- containerd: v2.2.2
- Kind: v0.32.0
- Kind Network: bridge, subnet
192.168.97.0/24, created 12 days prior
- Node Image: kindest/node:v1.36.1
Steps to Reproduce
- Create a Kind cluster with multiple nodes:
kind create cluster --name kind
- Deploy any pod that requires pulling an image from Docker Hub
- Observe
ErrImagePull / ImagePullBackOff with EOF errors
Diagnostic Results
DNS Resolution — ✅ Working
$ docker exec kind-worker getent ahosts registry-1.docker.io
44.207.198.110 STREAM registry-1.docker.io
18.213.142.16 STREAM
# ... (IPv4 addresses resolved correctly)
resolv.conf
$ docker exec kind-worker cat /etc/resolv.conf
nameserver 0.250.250.254
options ndots:0
# Based on host file: /etc/resolv.conf (internal resolver)
HTTPS Connectivity — ❌ Failed
$ docker exec kind-worker curl -sI --connect-timeout 5 https://registry-1.docker.io/v2/
# Exit code: 35 (SSL/TLS connection error)
$ docker exec kind-worker curl -sI --connect-timeout 5 https://google.com
# Exit code: 35
$ docker exec kind-worker curl -4 -sI --connect-timeout 5 https://registry-1.docker.io/v2/
# Exit code: 28 (Connection timeout)
Routing — Exists but no connectivity
$ docker exec kind-worker ip -4 route show default
default via 192.168.97.1 dev eth0
$ docker exec kind-worker ip -6 route show default
default via fc00:f853:ccd:e793::1 dev eth0 metric 1024 pref medium
Network Config
$ docker network inspect kind
# EnableIPv4: true, EnableIPv6: true
# Subnet: 192.168.97.0/24, Gateway: 192.168.97.1
# com.docker.network.bridge.enable_ip_masquerade: true
# MTU: 1500
Kubernetes Event Errors
Failed to pull image "redis:7-alpine": failed to resolve reference
"docker.io/library/redis:7-alpine": failed to do request:
Head "https://registry-1.docker.io/v2/library/redis/manifests/7-alpine": EOF
Expected Behavior
Kind nodes should be able to reach external HTTPS endpoints (Docker Hub, etc.) through OrbStack's NAT/bridge network, similar to how regular Docker containers can.
Notes
- Regular
docker pull from the host works fine
- System containers within Kind (
coredns, kindnet, kube-proxy) work because their images are pre-loaded
- The Kind network was created 12 days ago; the issue may be related to an OrbStack update or network state drift
- Workaround: Pre-loading images using
docker save | docker exec -i <node> ctr --namespace=k8s.io images import --digests -
Description
When running a Kind (Kubernetes in Docker) cluster on OrbStack, the Kind nodes are unable to reach any external HTTPS endpoint including Docker Hub (
registry-1.docker.io). This prevents the cluster from pulling container images at runtime.DNS resolution works correctly through OrbStack's internal resolver (
0.250.250.254), but TCP/HTTPS connections timeout or fail with TLS errors.Environment
192.168.97.0/24, created 12 days priorSteps to Reproduce
ErrImagePull/ImagePullBackOffwith EOF errorsDiagnostic Results
DNS Resolution — ✅ Working
resolv.conf
HTTPS Connectivity — ❌ Failed
Routing — Exists but no connectivity
Network Config
Kubernetes Event Errors
Expected Behavior
Kind nodes should be able to reach external HTTPS endpoints (Docker Hub, etc.) through OrbStack's NAT/bridge network, similar to how regular Docker containers can.
Notes
docker pullfrom the host works finecoredns,kindnet,kube-proxy) work because their images are pre-loadeddocker save | docker exec -i <node> ctr --namespace=k8s.io images import --digests -